Memory During Anesthesia

I was on an overnight transatlantic flight and said to my colleague, "The dinner cart is one cabin ahead. I'm going to take a benzodiazepine so that I get a good night's sleep after dinner." The following morning I awoke and said, "That worked quickly; I was asleep before dinner," to which he replied, "No, you were not. You ate dinner and we talked all through the meal." To this day, I have no recollection of that meal. Yet, I have no doubt that the memory exists somewhere in my subconscious. The same phenomenon may be true for memory of events during anesthesia. In 1965, Levinson [1] simulated a mock crisis during general anesthesia. During hypnosis, 1 mo after surgery, 4 out of 10 patients were able to reproduce the words spoken by the anesthesiologist and 4 other patients became anxious and awoke from hypnosis. This study suggested that patients may have indirect memory for events during general anesthesia without having actual recall. From this observation, a small industry has resulted, producing at least 200 articles and three international symposia, yet the phenomenon of indirect memory during anesthesia remains poorly understood. Many even doubt that the phenomenon exists. The literature is complicated because some of the studies have methodologic problems and other studies cannot be replicated [2-4]. The original Levinson study [1] was flawed and would not meet today's methodologic standards for publication. There was no control group, no randomization, and no blinding. An article by Chortkoff et al. [5] in this issue of Anesthesia & Analgesia attempts a replication of the Levinson experiment under laboratory conditions. This article should be considered against the background of other work in this field. A distinction has been drawn between explicit and implicit memory. Explicit memory requires the conscious recollection of a previous event, whereas implicit memory is shown by a change in task performance, attributable to information acquired, but without necessarily having a direct recollection of the event [6]. For example, if a tape was played to a patient during surgery instructing that patient to pull on his or her ear in the postoperative interview, and the patient indeed performed this task, yet without recalling the content of the tape, then this would be evidence of indirect memory during anesthesia. Such a study has been performed with positive results [7]. However, no attempt was made to control the quantity of anesthetic administered, and the tape containing the suggestion was played approximately 5 min before reversal of anesthesia, when anesthesia may have been deliberately "light." Thus, positive results from this study are not surprising. A variety of word recognition tasks have been used to investigate the phenomenon of indirect memory during anesthesia, often with conflicting results. Kihlstrom et al. [8] presented paired word associates (e.g., foot-hand), then tested for free recall and cued recall of the word pairs and also conducted a free association test immediately after surgery and 2 wk postoperatively (by telephone). Anesthesia was maintained with isoflurane 1.0%-1.4% end tidal concentration in oxygen. Both free and cued recall tasks gave negative results at both testing times, yet the free association test showed evidence of implicit memory at both immediate and delayed testing. The same group repeated this study with sufentanil/nitrous oxide anesthesia [9] with completely negative results. Intuitively, one would expect the opposite result, since sufentanil/nitrous oxide would not usually be considered as "complete" an anesthetic [10] as isoflurane. Many other studies of this type have been published, with nearly equal numbers of positive and negative results [see the reviews [11,12] and the books of conference proceedings [13,14]]. It is possible that some of the material presented in the studies described above was not processed and recalled because it was irrelevant to the patient. There is evidence suggestive of selective auditory information processing during sleep. If my child cries at night, my wife wakes and I do not. If my beeper activates, I am immediately awake and my wife is not. To paraphrase the example by Levinson [15], I am walking along a rope bridge across a chasm. A thousand feet below me is a raging river. Falling means certain death. My whole being is focused on reaching the other side. Behind me someone is saying, "table, chair, tug on your ear." Imagine how much better it would be if the voice were saying, "Everything is going fine. You are doing very well." The effect of presenting such material to patients in the form of positive therapeutic suggestions has been extensively investigated [11]. Such suggestions have been investigated in a double-blind study in women undergoing hysterectomy [2]. The patients in the suggestion group spent significantly less time in hospital and had a shorter duration of pyrexia than the controls. However, other investigators attempted to replicate this study and had negative results [3]. It has been suggested that these conflicting results are due to a different distribution of data in one control group, resulting from chance bias in allocation of patients to that group [4]. Positive results in terms of decreased hospital stay have also been reported in patients undergoing cholecystectomy [16]. Yet the same investigators were unable to obtain the same effect in a subsequent study using the same surgical procedure [17]. There are many factors contributing to postoperative hospital stay, and it would be surprising if positive therapeutic suggestion had a consistent effect on hospital stay. Furthermore, duration of hospital stay is dependent on many factors outside the control of such a study. Patient-controlled analgesia dose represents an observer-independent measure of the effect of positive therapeutic suggestions. It has been demonstrated, in two double-blind studies by different groups of investigators, that patient-controlled analgesia dose requirements were reduced in patients who received positive therapeutic suggestions [18,19]. These studies provide strong evidence for auditory information processing of material relevant to the well-being of the patient. If positive therapeutic suggestion is associated with a beneficial outcome, then negative suggestion should result in an adverse outcome, yet this hypothesis is difficult, if not impossible, to test ethically in patients. The article by Chortkoff et al. [5] explores this problem in volunteers under carefully controlled laboratory conditions. The oxygen failure crisis used by Levinson [1] was reproduced in a double-blind manner but no evidence of implicit memory of the crisis was found with either propofol or desflurane, nor was there evidence of learning of matter-of-fact information. Does this study provide definitive evidence that, with adequate anesthesia, auditory information processing does not occur during surgical anesthesia? I think the answer is no for several reasons. Differences in study protocol, informed consent, anesthetic technique (either versus propofol and desflurane) and measurement time interval may have relevance to the failure to replicate previous findings [1]. Anecdotal evidence is presented that two of the volunteers had recall of the crisis (although three volunteers gave similar information after the control situation). One investigator, when he was confident of his choice, was accurate in distinguishing the post crisis interview, although his overall success approximated that expected by chance. Differences in study populations may be of major importance. Volunteers entering a study for which they will be paid have a different expectation than patients undergoing elective surgery. The volunteer should expect that the outcome will be satisfactory and no adverse events will occur (even given the detailed consent), whereas a surgical patient may be very anxious, fearing cancer or even death. The volunteers were not subjected to the pain of surgery, and this may have influenced learning. The differences between patients and volunteers may also be related to differing metabolic responses. It has been suggested that hormonal responses, particularly epinephrine, may increase the ability to learn [20], although this issue is controversial [21]. A large body of literature also suggests that patients do process auditory information during adequate anesthesia. Defining adequate anesthesia in surgical patients remains difficult. Implicit memory for neutral content information does not occur at 0.6 minimum alveolar anesthetic concentration isoflurane in unstimulated patients or at 1.0 minimum alveolar anesthetic concentration isoflurane during surgery [22]. However, the lower limit of anesthetic concentration at which emotionally charged information is registered is unknown, and we currently lack a generally accepted direct measure of adequacy of anesthesia. While variations in adequacy of anesthesia and variations in study populations may account for many of the conflicting outcomes in the literature, other relevant factors include subtle details of experimental design as well as the exact tasks that are studied. If indirect memory exists during adequate general anesthesia, it is a fragile phenomenon. Since so much research has failed to give clear indications as to when the phenomenon occurs, we are unlikely to see a resolution to the issue in the next few years. With our current knowledge, assuming that patients may process auditory information during general anesthesia, the practical implications are obvious. Inappropriate remarks about the patient, their body habitus, or pathology should never be made and proper decorum should always be observed in the operating room. Talking to the anesthetized patient and telling him or her that everything is going well is reasonable. However, if a real, rather than a simulated drama occurs, remember that your patient may be listening.